CN1873660A - Supervised classification process of artificial immunity in remote sensing images - Google Patents
Supervised classification process of artificial immunity in remote sensing images Download PDFInfo
- Publication number
- CN1873660A CN1873660A CN 200610019506 CN200610019506A CN1873660A CN 1873660 A CN1873660 A CN 1873660A CN 200610019506 CN200610019506 CN 200610019506 CN 200610019506 A CN200610019506 A CN 200610019506A CN 1873660 A CN1873660 A CN 1873660A
- Authority
- CN
- China
- Prior art keywords
- antibody
- sample
- memory
- artificial immunity
- artificial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a classification method that monitor remote sensing video by manual immunity, its characteristic are:(1) opening remote sensing video;(2) choosing zone, input parameter; (3) calculating saturation value, choose many kind of original manual identified group and original antibody memory bank; (4) training all original stylebook in the stylebook array used manual immunity system;(5) selecting next stylebook, repeat (3)-(4) until all stylebooks are trained, and receiving all zone antibody memory bank; (6) to whole video, comparing with distance between image cell and antibody memory bank, classifying the image cell to category whose distance between antibody memory bank and the image is minimal.
Description
Technical field
The invention belongs to the remote sensing image processing technology field, especially a kind of remote sensing image supervised classification method based on artificial immune system.
Background technology
The thought of remote sensing image supervised classification is according to the known sample class and the priori of classification, determine discriminant function and corresponding criterion, then with the observed reading substitution discriminant function of the sample of unknown classification, make a determination according to the affiliated classification of criterion again, realize extracting based on the geography information of remotely-sensed data to this sample.Traditional remote sensing image supervised classification method mainly comprises: parallel pipe Dow process, minimum distance method, maximum likelihood method.
The parallel pipe Dow process supposes that based on the spectral pattern of atural object the spectral pattern of similar atural object is similar as the standard of differentiating.A classification range of variation is set, comes if pixel is within the classification range of variation, then it to be included into the classification at place unknown pixel classification according to this classification range of variation or judgement district.If pixel is outside all categories range of variation, stipulate that then it is unknown pixel.This method is calculated simple, but exists under the situation of correlativity, and the effect of the capable determinating area match of square classification training data is very bad, causes wrong branch and mistake to divide easily.
In order to improve nicety of grading, utilize distance discrimination function and decision rule to propose minimum distance method.Minimum distance method is to utilize the average at each wave band of all categories in the training sample, determines its classification according to each pixel to the size of training sample mean distance.The method of this anomaly average minimum distance classification is easy on mathematics, and calculating also is effectively, but still there is certain limitation in it, and particularly it is to there being the spectral response data that change in various degree also insensitive.
Maximum likelihood method is classical sorting technique, is widely used in classification of remote-sensing images.It mainly specifies the classification of each pixel according to similar spectral quality and the hypothesis that belongs to the probability maximum of certain class.The total misclassification probability of maximum likelihood method is less than the total misclassification probability of minimum distance method, is normal state but this method is supposition atural object all kinds of distribution function, can not guarantee then that when all kinds of distributions of atural object are not normal state nicety of grading reaches requirement.
Relevant document: Sun Jiabing, oxazepan, Guan Zequn.Remote sensing principle methods and applications [M]. Beijing: Mapping Press, 1997; Soup Guoan, Zhang Youshun, Liu Yongmei. process in remote sensing digital image processing etc. [M]. Beijing: Science Press, 2004; Zhao's inch. remote sensing application analysis principle and method [M]. Beijing: Science Press, 2003; Campbell, J.B., Introduction to Remote Sensing[M] .London:Taylor ﹠amp; Francis, 2002.
In sum, maximum likelihood method in traditional remote sensing image supervised classification method needs in advance that supposition atural object is all kinds of to be normal distribution, and this condition differs and is met surely in actual atural object distributes; And parallel pipe Dow process and minimum distance method all are to adopt the average of all samples of sampling one's respective area to obtain to the calculating of cluster centre, and therefore resulting cluster centre has locality when training sample, and fails to consider the of overall importance of sample.
Artificial immune system (Artificial Immune System, be called for short AIS) is to be subjected to the inspiration of Immune System and a kind of novel intelligence computation method that produces.In the past few years, the application of AIS has expanded to numerous areas such as information security, pattern-recognition, machine learning, data mining gradually, demonstrates powerful information processing of AIS and problem solving ability and wide research and application prospect.Relevant document: D.Dasgupta, Artificial Immune Systems and Their Applications, Germany:Springer, 1999; L.N.de Castro and J.Timmis, Artificial Immune systems:A New Computational Intelligence Approach, London, U.K.:Springer-Verlag, 2002; J.Timmis, M.Neal, and J.E.Hunt, " An artificial immune systemfor data analysis, " Biosystem, 55 (1/3), 2000; Xiao Renbin, Wang Lei. artificial immune system: principle, model, analysis and prospect [J]. Chinese journal of computers, 2002,25 (12).
AIS is the very strong optimisation technique of a kind of self-adaptation, numerous attributes of Immune System have been inherited, have self-organization, self study, self-identifying, the ability of memory certainly, therefore it can provide 90% the hunting zone that reaches optimum solution fast, thereby can obtain globally optimal solution comparatively fast more accurately, this be other optimisation technique can't be obtained.Relevant document: L.N.De Castro and F.J.Von Zuben, " Learning andoptimization using the clonal selection principle; " IEEE Trans.onEvolutionary Computation, Vol.6 (3): 2002; Atkinson P M, Lewis P.Geostatistical classification for remote sensing:an introduction[J] .Computers ﹠amp; Geosciences, 26,2000; Adams D.How the immune system worksand why it causes autoimmune diseases[J] .Immunology Today, 17 (7), 1996.
Yet in the remote sensing image supervised classification, AIS is not also well used.How using for reference Immune System, thereby the supervised classification process of artificial immunity of high-class precision is provided, is the technical matters that present remote sensing technology field needs to be resolved hurrily.
Summary of the invention
The present invention seeks to utilize the advantage of artificial immune system, a kind of supervised classification method that is used to improve the classification of remote-sensing images precision is provided.
For achieving the above object, the invention provides the supervised classification process of artificial immunity of remote sensing image:
(1) opens remote sensing image to be classified by the remote sensing image handling procedure;
(2) distribute and the class categories number according to actual atural object, on remote sensing image, utilize sample district instrument to select sample interested district or training field, training sample is saved as the sample array, the input algorithm parameter;
(3) utilize all sample antigens to calculate the affinity threshold value, select sample antigen then at random, obtain all kinds of original manual identified ball populations and original antibody memory bank, deposit all kinds of artificial immunity identification nodule number group and memory antibody array in;
(4) all the antigen samples in the sample array are carried out artificial immune system training, obtain the antibody memory bank in all sample districts, the training of each antigen sample is comprised following 5 steps;
(4.1) calculate the irritation level of sample antigen each memory antibody in the similar original antibody memory bank, from similar original antibody memory bank, find the memory antibody that mates most with this antigen;
(4.2) clone mating most memory antibody, obtain the clonal antibody array, each clonal antibody in the clonal antibody array is made a variation, deposit the antibody after the variation in similar artificial immunity identification ball population array;
(4.3) artificial immunity identification ball population is suppressed to handle;
(4.4) for the artificial immunity identification ball population after handling through step (4.3), whether the average irritation level of judging this population satisfies the setting irritation level, if satisfy then enter step (4.5), do not satisfy then population is carried out the clonal vaviation operation, the new population behind the clonal vaviation is begun to recomputate processing up to satisfying threshold condition by step (4.3);
(4.5) from artificial immunity identification ball population, obtain cell with sample antigenic stimulus level maximum as candidate's memory antibody, judge the irritation level size of the coupling memory antibody that obtains in candidate's memory antibody and the step (4.1) then, if greater than coupling memory cell irritation level, then candidate's memory antibody enters in the antibody memory bank, deposit the memory antibody array in, and then calculating affinity between the two, if satisfying the affinity threshold values of step (3) calculating then mate antibody, affinity from the memory antibody array, removes;
(5) select next training sample, repeating step (3) up to the sample training of finishing all sample districts, obtains the antibody memory bank in all sample districts to step (4);
(6) to the view picture image, the distance of each pixel memory antibody in the antibody memory bank is relatively adjudicated this pixel in the classification under that minimum memory antibody of distance.
And, algorithm parameter includes cloning efficiency, irritation level threshold value and system resource threshold value, clone mating most memory antibody according to cloning efficiency, whether the average irritation level of discerning the ball population according to the artificial immunity of irritation level threshold decision satisfies the setting irritation level.
And, adopt the resource limit method that artificial immunity identification ball population is suppressed to handle, may further comprise the steps, the irritation level of at first calculating each artificial immunity identification ball in the artificial immunity identification ball population column criterionization of going forward side by side is calculated the shared artificial immune system resource of each artificial immunity identification ball according to irritation level; Calculate the total resources of artificial immunity identification ball population then, if total resources surpass the system resource threshold values, the artificial immunity identification ball that then deducts the irritation level minimum allows resource up to total resources smaller or equal to system.
And, carrying out step (2) afterwards, the proper vector of standardization sample antigen is to guarantee that the value of sample antigen to the distance of antibody or antibody to the distance between the antibody is between 0～1.
Characteristics of the present invention are: by the training of artificial immune system to sample district sample, showed the distribution situation of sample more accurately, avoided the distribution locality; Obtaining on the sample distribution basis, evolving by artificial immune system obtains the cluster centre or the memory antibody of each classification, has avoided adopting each sample district average of simple computation and the limitation that obtains cluster centre.The present invention also adopts the resource limit method to reduce the redundancy of artificial immune system, accelerates convergence of algorithm speed, reduces the classification time of remote sensing image; By cluster centre or memory antibody that evolution obtains unknown pixel is classified, improved nicety of grading.The intelligent height of the inventive method is carried out the efficient height, is applicable to multispectral, target in hyperspectral remotely sensed image classification, can effectively improve the nicety of grading of remote sensing image.
Description of drawings
Fig. 1 form spatial model;
Fig. 2 Immune System and artificial immune system corresponding diagram;
Fig. 3 embodiment of the invention principal function flow chart;
Fig. 4 embodiment of the invention initialization function program block diagram;
Fig. 5 embodiment of the invention ARB produces the function program block diagram;
Fig. 6 embodiment of the invention ARB resource contention function program block diagram;
Fig. 7 embodiment of the invention data base evolution function program block diagram.
Embodiment
For the ease of understanding the present invention, at first provide theoretical foundation of the present invention:
One of immune critical function is to remove external foreign matter by producing antibody (antibody), and foreign matter can be microorganism (bacterium, virus etc.), special-shaped haemocyte, grafting device official rank, and they all are called antigen (antigen).Immune basic composition is lymphocyte or white blood cell.These special cells mainly can be divided into B cell and T cell two big classes.These two kinds of cells all have own unique ecologic structure and produce many Y type antibody from their surface and kill antigen.
In order to describe the interaction between antibody and the antigen quantitatively, Perelson and Oster proposed form space (shape-space) model in 1979.The form spatial model has been described the combination degree between the antigen of antibody and combination with it.As shown in Figure 1, it is the zone of V that a volume is arranged in the form space S, wherein contains paratope (with representing) and epitope (usefulness * expression) shape complementarity zone.Suppose that wherein an antibody can discern all volume V around it
εComplementary epitope in the scope, therefore in immune system, a limited number of antibody can be discerned the epitope of infinite number.
Form spatial model according to Perelson is introduced new ideas---and the artificial cognition ball (ArtificialRecognition Ball, ARB).The B cell that ARB is used for describing having congruence property in a large number, its purpose mainly are in order to reduce a large amount of clonal antibodies and the antibody through surviving behind the clonal vaviation that limits in the population in the antibody population.In order to understand better, the invention provides accompanying drawing 2, described the corresponding relation of Immune System and artificial immune system: antibody is corresponding to cluster centre or proper vector; The corresponding training sample data of antigen; Identification ball correspondence has the mixture of proper vector and vectorial class; The span of form space character pair vector; The corresponding optimum artificial cognition ball clone of clonal expansion; Ripe corresponding artificial cognition ball variation of affinity and the low artificial cognition ball of minimizing irritation level; The corresponding antibody memory bank of immunological memory; System dynamics balance correspondence is removed the low artificial immunity identification ball of irritation level continuously, is produced new artificial immunity identification ball and memory antibody.
Based on this Immune Clone Selection theory, the invention provides the supervised classification process of artificial immunity of remote sensing image, remote sensing image processing data complexity, workload is big, generally need to adopt computer means to realize, therefore technical solution of the present invention has adopted computer program and computerese to be described, and for example the memory antibody array comes down in order to explain the memory antibody set that a content changes according to evolution.The claimed technical scheme of the present invention is not limited to the computer program flow process, and should comprise that other are equal to the replacement means.Technical solution of the present invention is as follows:
(1) opens remote sensing image to be classified by the remote sensing image handling procedure;
(2) distribute and the class categories number according to actual atural object, on remote sensing image, utilize sample district instrument to select sample interested district or training field, training sample is saved as the sample array, the input algorithm parameter;
(3) utilize all sample antigens to calculate the affinity threshold value, select sample antigen then at random, obtain all kinds of original manual identified ball populations and original antibody memory bank, deposit all kinds of artificial immunity identification nodule number group and memory antibody array in;
(4) all the antigen samples in the sample array are carried out artificial immune system training, obtain the antibody memory bank in all sample districts, the training of each antigen sample is comprised following 5 steps;
(4.1) calculate the irritation level of sample antigen each memory antibody in the similar original antibody memory bank, from similar original antibody memory bank, find the memory antibody that mates most with this antigen;
(4.2) clone mating most memory antibody, obtain the clonal antibody array, each clonal antibody in the clonal antibody array is made a variation, deposit the antibody after the variation in similar artificial immunity identification ball population array;
(4.3) artificial immunity identification ball population is suppressed to handle;
(4.4) for the artificial immunity identification ball population after handling through step (4.3), whether the average irritation level of judging this population satisfies the setting irritation level, if satisfy then enter step (4.5), do not satisfy then population is carried out the clonal vaviation operation, the new population behind the clonal vaviation is begun to recomputate processing up to satisfying threshold condition by step (4.3);
(4.5) from artificial immunity identification ball population, obtain cell with sample antigenic stimulus level maximum as candidate's memory antibody, judge the irritation level size of the coupling memory antibody that obtains in candidate's memory antibody and the step (4.1) then, if greater than coupling memory cell irritation level, then candidate's memory antibody enters in the antibody memory bank, deposit the memory antibody array in, and then calculating affinity between the two, if satisfying the affinity threshold values of step (3) calculating then mate antibody, affinity from the memory antibody array, removes;
(5) select next training sample, repeating step (3) up to the sample training of finishing all sample districts, obtains the antibody memory bank in all sample districts to step (4);
(6) to the view picture image, the distance of each pixel memory antibody in the antibody memory bank is relatively adjudicated this pixel in the classification under that minimum memory antibody of distance.
The present invention has designed computer program and has realized supervised classification task of the present invention, program adopts supervised classification process of artificial immunity that image antigen I mageAg is trained, if arrive end condition then classify and finish, otherwise loop iteration training image antigen satisfies up to end condition.In order to make program succinct, and convenient the combination with existing remote sensing image process software and the program realization, the function call thinking adopted in large quantities.
In order to improve the classification effectiveness of this method, the invention provides the resource limit method, may further comprise the steps: the irritation level of at first calculating each artificial immunity identification ball in such artificial immunity identification ball population column criterionization of going forward side by side, calculate the shared artificial immune system resource of each artificial immunity identification ball according to irritation level; Calculate the total resources of such artificial immunity identification ball population A RB then, allow resource if total resources surpass system, the artificial immunity identification ball that then deducts the irritation level minimum allows resource up to total resources smaller or equal to system.When artificial immune system was used to solve all kinds of engineering problem, core problem was how to upgrade antibody memory bank the most.Because the renewal of memory cell is obtained through stimulation oversaturation and inhibition by general antibody in the antibody memory bank, so the stimulation of antibody and method that inhibition the is adopted gordian technique that just become the using artificial immune system to deal with problems.Based on above theory, the present invention has designed the evolution thinking of antibody memory bank: in the present invention, adopt the resource limit method to come the ARB population is stimulated and inhibition, thereby reach the purpose of evolution ARB population data base.Its detailed process is to certain invasion antigen A g, calculate the irritation level of Ag to each ARB antibody, carry out resources allocation according to its irritation level according to resource allocation mechanism, irritation level is high more, ARB just can have more resources and B cell, and wherein the total resources of ARB population are fixed.After the experience resources allocation,, think then that this ARB can not represent training data again and from network sweep if ARB loses all B cells or has low irritation level.Calculate the average irritation level of ARB population,, then finish training this antigen if its average irritation level reaches given stimulus threshold, otherwise evolution ARB population and Resources allocation again.Experimental results show that the resource limit method had both guaranteed the antibody diversity in the renewal of ARB antibody population, had avoided the prematurity convergence again simultaneously and had accelerated convergence of algorithm speed.
Calculate for convenience, the present invention is also carrying out step (2) afterwards, and the proper vector of standardization sample antigen is to guarantee that the value of sample antigen to the distance of antibody or antibody to the distance between the antibody is between 0～1.Carry out normalization and brought facility to calculating aberration rate afterwards.
Describe technical solution of the present invention in detail below in conjunction with the concrete implementation step of embodiment:
(1) utilize remote sensing image processing system, by input image width, highly, wave band number and data type open the input remote sensing image.Total wave band number of remote sensing image adopts N in the embodiment of the invention
bMark.This process belongs to the image input process, and the realization program is not introduced in detail.At present, the computer utility in remote sensing field is very general, it is conventional means that the digital picture that adopts the remote sensing image handling procedure that remote sensing is obtained is handled, and has possessed the remote sensing image handling procedure of opening image function and sampling instrument substantially and all can use for the invention process.During concrete enforcement, can be combined into one other step required functions of the basic function of remote sensing image handling procedure and the inventive method, can realize by computer programming.
(2) atural object according to reality distributes and required class categories number, utilizes sample district instrument to select sample interested district on remote sensing image, opens up a sample array TrainAg in computing machine, deposits training sample in the sample array.This array type is a structure ROIAg type, and structure comprises sample data and two structure variablees of sample data classification.Can be provided with by program during concrete enforcement the algorithm parameter input frame is provided, behind the ejection algorithm parameter input frame, the required parameter of input algorithm mainly comprises: cloning efficiency R
Clone, irritation level threshold value S
Threshold, resource threshold E
FairSet up the executive routine that activates this algorithm behind the algorithm parameter.Program employing artificial immune system is trained the sample antigen A g in sample district, if training is finished then classify, finishes up to all sample antigens training otherwise circulate.The principal function program flow chart is seen accompanying drawing 3.
Technical scheme provided by the invention also standardization the proper vector of sample antigen, to guarantee that the value of sample antigen to the distance of antibody or antibody to the distance between the antibody is between 0 ~ 1.Specific embodiment calls normalization function Normalization () then and realizes by entering initialization function Initialization () inlet.
(3) utilize among the sample array TrainAg all sample antigens to calculate affinity threshold values (affinitythreshold is labeled as AT), this threshold value will be in step 4.5 plays a key effect during the evolution data base.Suppose that certain training sample set contains n antigen (antigen is labeled as g), then AT is the average affinity of this training sample set, and its computing formula is as follows:
Wherein the affinity computing formula is
In the following formula formula, g
iI the antigen that the expression training sample is concentrated, wherein g
IkExpression antigen g
iThe numerical value of k wave band.
Concentrate the antigen of selecting some at random that its numerical value assignment is discerned ball and certain memory antibody to artificial immunity from training sample then, obtain original manual identified ball population A B and original antibody memory bank MC, deposit artificial cognition ball population array ABArray and antibody memory bank array MCarray respectively in.The specific embodiment step is referring to accompanying drawing 4, and normalization function Normalization () calculates AT after calculating and finishing; Select to obtain initial ABArray content and initial MCArray content at random; Call ARB then and produce function G enerateARB (), to enter next step.
(4) all the antigen samples among the sample array TrainAg are carried out artificial immune system training, obtain the antibody memory bank in all sample districts, the training of each antigen sample is comprised following 5 steps;
(4.1) for certain sample antigen g, from the data base similar, find the memory cell C that mates most with this antigen with this antigen
Join, wherein
Wherein stimulation (x, y)=(x, y), maximal value is got in the argmax symbolic representation to 1-affinity.
(4.2) work as C
JoinAfter determining, to C
JoinClone, obtain clonal antibody array CloneAB, traversal clonal antibody array CloneAB, each clonal antibody is made a variation with corresponding aberration rate, the new artificial immunity identification ball that obtains after the variation is deposited among the artificial immunity identification ball population array ABArray of corresponding classification, wherein clone number N
CloneFor: N
Clone=R
Hyper* N
Clone* stimulation (g, C
Join), R wherein
HyperRepresent the different value of immune hypermutation, general value is 2.Referring to Fig. 5, the embodiment of the invention adopts the mode of calling variation function mutate () to realize mutation operation, and function thes contents are as follows: certain clonal antibody among the clonal antibody array CloneAB, ARB
j, carrying out following operation, variable j of every operation is from adding 1, up to the total N of variable j＞clone
CloneIn time, finish to clonal antibody array CloneAB variation; For ARB
jI wave band value vi, minimum value minvi by calculating i wave band (assignment is given variable ai) and maximal value maxvi (assignment is to variable bi) make a variation by following formula:
Random in the following formula (1,1) random number functions produces the random number of scope in [1,1] by this function.Specific embodiment is given variable rd_mr with random number random (0,1) assignment, gives variable rd_mr with random number random (1,1) assignment, so that computing machine carries out computing.
(t, the formula that embodies y) can be taken as the function Δ
Wherein r is a random number on [0,1], and T is the maximum algebraically of variation, and t is current variation algebraically, and λ is a parameter of decision nonuniformity degree, and it plays a part to adjust the Local Search zone, and its value is generally 2 to 5.
(4.3) adopt the resource limit method that artificial immunity identification ball population is handled.Referring to accompanying drawing 6, realize by ARB resource contention function ARBRescomep () in embodiments of the present invention.Concrete steps are as follows:
At first artificial immunity is discerned the irritation level of ball ARB and is calculated its resource among the standardization artificial immunity identification ball population array ABArray.At first in artificial immunity identification ball population array ABArray, obtain its highest irritation level (S
Max) and minimum irritation level (S
Min); Each artificial immunity identification ball ARB among the identification of the artificial immunity herein ball population array ABArray is labeled as b, for each b ∈ ABArray, and its irritation level of standardization, the standardization formula is
For each b ∈ ABArray, calculate its resource b
Resource=b
Stim* R
CloneCorresponding steps among the ARB resource contention function ARBRescomep () is: input artificial immunity identification ball population array ABArray content; Obtain the highest irritation level (S of artificial immunity identification ball by the irritation level of adding up all artificial immunity identification balls
Max) and minimum irritation level (S
Min); The irritation level of each artificial cognition ball b in the standardized A BArray array; The resource of each artificial cognition ball in the normalized ABArray array.
Add up all resources of ARB among the ABArray then, obtain total resources E
Always, compare E
AlwaysWith the resource threshold E that allows
FairSize, if E
AlwaysGreater than E
Fair, ball is discerned in the artificial immunity that then at first deducts the irritation level minimum, up to satisfying threshold value E
FairTill.
(4.4) judge that whether the average irritation level s of this population is greater than irritation level threshold value S
ThresholdIf, enter step (4.5), not satisfy then population is carried out the new artificial immunity identification ball population of clonal vaviation operation generation, this clonal vaviation operation can refer step (4.2).New population behind the clonal vaviation is begun to recomputate processing up to satisfying threshold condition by step (4.3).
(4.5) calculate the irritation level of antigen to all b ∈ ABArray, the b that irritation level is the highest elects candidate's memory cell C as
WaitJudge that antigen g is to C
WaitAnd C
JoinIrritation level, if stimulation (g, C
Wait) greater than stimulation (g, C
Join) then with C
WaitAdd among the data base MCArray.And then calculating C
WaitAnd C
JoinBetween affinity, if affinity affinity (C
Wait, C
Join) less than the product of affinity threshold value A T and affinity threshold percentage ATS then with C
JoinFrom data base MCArray, remove, to guarantee the memory cell (ATS sets in advance, in order to provide setting range) within certain quantity among the data base MCArray.Referring to Fig. 7, the embodiment of the invention realizes above operation by calling data base evolution function DevelopMCpop (), and variable aff is used for mark affinity affinity (C among the figure
Wait, C
Join), S (g, C
Wait) be used for mark stimulation (g, C
Wait), S (g, C
Join) be used for mark stimulation (g, C
Join).
Carry out step (4.1) to step (4.6) for all the antigen samples in the sample array, can finish such sample training process.
(5) select next training sample, repeating step (3) up to the sample antigen training of finishing selected all the sample districts of sample district instrument in step (2), obtains the antibody memory bank in all sample districts to step (4).
(6) to the view picture image, the distance of each pixel memory antibody in the antibody memory bank is relatively adjudicated this pixel in the classification under that minimum memory antibody of distance.
After adopting function call thought to be technical scheme programming of the present invention, whole program is implemented structure and is: import remote sensing image to be classified; Select the sample district; Definition sample array TrainAg; Enter the principal function inlet; Algorithm parameter is set; Calling classification function AISClassifier (); After entering classification function AISClassifier () inlet, the sample district data among the input sample array TrainAg; Call initialization function Initialization (), determine initial antibodies population and memory antibody; Call ARB and produce function G enerateARB (); Call ARB resource contention function ARBRescomep (); Whether the average irritation level s that judges this population is greater than irritation level threshold value S
Threshold, otherwise return ARB resource contention function ARBRescomep (); If average irritation level s is greater than irritation level threshold value S
Threshold, call data base evolution function DevelopMCPop (); Select next sample, all finished by training up to all sample district antigens; Finish last classification judgement, obtain the classification results image.
Claims (4)
1. the supervised classification process of artificial immunity of a remote sensing image is characterized in that:
(1) opens remote sensing image to be classified by the remote sensing image handling procedure;
(2) distribute and the class categories number according to actual atural object, on remote sensing image, utilize sample district instrument to select sample interested district or training field, training sample is saved as the sample array, the input algorithm parameter;
(3) utilize all sample antigens to calculate the affinity threshold value, select sample antigen then at random, obtain all kinds of original manual identified ball populations and original antibody memory bank, deposit all kinds of artificial immunity identification nodule number group and memory antibody array in;
(4) all the antigen samples in the sample array are carried out artificial immune system training, obtain the antibody memory bank in all sample districts, the training of each antigen sample is comprised following 5 steps;
(4.1) calculate the irritation level of sample antigen each memory antibody in the similar original antibody memory bank, from similar original antibody memory bank, find the memory antibody that mates most with this antigen;
(4.2) clone mating most memory antibody, obtain the clonal antibody array, each clonal antibody in the clonal antibody array is made a variation, deposit the antibody after the variation in similar artificial immunity identification ball population array;
(4.3) artificial immunity identification ball population is suppressed to handle;
(4.4) for the artificial immunity identification ball population after handling through step (4.3), whether the average irritation level of judging this population satisfies the setting irritation level, if satisfy then enter step (4.5), do not satisfy then population is carried out the clonal vaviation operation, the new population behind the clonal vaviation is begun to recomputate processing up to satisfying threshold condition by step (4.3);
(4.5) from artificial immunity identification ball population, obtain cell with sample antigenic stimulus level maximum as candidate's memory antibody, judge the irritation level size of the coupling memory antibody that obtains in candidate's memory antibody and the step (4.1) then, if greater than coupling memory cell irritation level, then candidate's memory antibody enters in the antibody memory bank, deposit the memory antibody array in, and then calculating affinity between the two, if satisfying the affinity threshold values of step (3) calculating then mate antibody, affinity from the memory antibody array, removes;
(5) select next training sample, repeating step (3) up to the sample training of finishing all sample districts, obtains the antibody memory bank in all sample districts to step (4);
(6) to the view picture image, the distance of each pixel memory antibody in the antibody memory bank is relatively adjudicated this pixel in the classification under that minimum memory antibody of distance.
2. supervised classification process of artificial immunity as claimed in claim 1, it is characterized in that: algorithm parameter includes cloning efficiency, irritation level threshold value and system resource threshold value, clone mating most memory antibody according to cloning efficiency, whether the average irritation level of discerning the ball population according to the artificial immunity of irritation level threshold decision satisfies the setting irritation level.
3. supervised classification process of artificial immunity as claimed in claim 1, it is characterized in that: adopt the resource limit method that artificial immunity identification ball population is suppressed to handle, may further comprise the steps, the irritation level of at first calculating each artificial immunity identification ball in the artificial immunity identification ball population column criterionization of going forward side by side is calculated the shared artificial immune system resource of each artificial immunity identification ball according to irritation level; Calculate the total resources of artificial immunity identification ball population then, if total resources surpass the system resource threshold values, the artificial immunity identification ball that then deducts the irritation level minimum allows resource up to total resources smaller or equal to system.
4. as claim 1 or 2 or 3 described supervised classification process of artificial immunity, it is characterized in that: carrying out step (2) afterwards, the proper vector of standardization sample antigen is to guarantee that the value of sample antigen to the distance of antibody or antibody to the distance between the antibody is between 0～1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100195062A CN100380395C (en) | 2006-06-29 | 2006-06-29 | Supervised classification process of artificial immunity in remote sensing images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100195062A CN100380395C (en) | 2006-06-29 | 2006-06-29 | Supervised classification process of artificial immunity in remote sensing images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1873660A true CN1873660A (en) | 2006-12-06 |
| CN100380395C CN100380395C (en) | 2008-04-09 |
Family
ID=37484131
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2006100195062A Expired - Fee Related CN100380395C (en) | 2006-06-29 | 2006-06-29 | Supervised classification process of artificial immunity in remote sensing images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100380395C (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859328A (en) * | 2010-06-21 | 2010-10-13 | 哈尔滨工程大学 | Exploitation method of remote sensing image association rule based on artificial immune network |
| CN105160291A (en) * | 2015-07-02 | 2015-12-16 | 上海闻泰电子科技有限公司 | Age estimation system based on artificial immunization identification system and method |
| CN112991357A (en) * | 2019-12-18 | 2021-06-18 | 中国船舶重工集团公司第七一一研究所 | Image segmentation method, system, computer device, readable storage medium and ship |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100595782C (en) * | 2008-04-17 | 2010-03-24 | 中国科学院地理科学与资源研究所 | Classification method for syncretizing optical spectrum information and multi-point simulation space information |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6018587A (en) * | 1991-02-21 | 2000-01-25 | Applied Spectral Imaging Ltd. | Method for remote sensing analysis be decorrelation statistical analysis and hardware therefor |
| EP1191459A1 (en) * | 2000-09-22 | 2002-03-27 | Nightingale Technologies Ltd. | Data clustering methods and applications |
| US20030026484A1 (en) * | 2001-04-27 | 2003-02-06 | O'neill Mark | Automated image identification system |
| CN1790379A (en) * | 2004-12-17 | 2006-06-21 | 中国林业科学研究院资源信息研究所 | Remote sensing image decision tree classification method and system |
-
2006
- 2006-06-29 CN CNB2006100195062A patent/CN100380395C/en not_active Expired - Fee Related
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859328A (en) * | 2010-06-21 | 2010-10-13 | 哈尔滨工程大学 | Exploitation method of remote sensing image association rule based on artificial immune network |
| CN101859328B (en) * | 2010-06-21 | 2012-07-11 | 哈尔滨工程大学 | Exploitation method of remote sensing image association rule based on artificial immune network |
| CN105160291A (en) * | 2015-07-02 | 2015-12-16 | 上海闻泰电子科技有限公司 | Age estimation system based on artificial immunization identification system and method |
| CN112991357A (en) * | 2019-12-18 | 2021-06-18 | 中国船舶重工集团公司第七一一研究所 | Image segmentation method, system, computer device, readable storage medium and ship |
| CN112991357B (en) * | 2019-12-18 | 2023-04-18 | 中国船舶集团有限公司第七一一研究所 | Image segmentation method, system, computer device, readable storage medium and ship |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100380395C (en) | 2008-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112308158B (en) | Multi-source field self-adaptive model and method based on partial feature alignment | |
| Hernández-Lobato et al. | Parallel and distributed Thompson sampling for large-scale accelerated exploration of chemical space | |
| Young et al. | Optimizing deep learning hyper-parameters through an evolutionary algorithm | |
| Almeida et al. | Playing tag with ANN: boosted top identification with pattern recognition | |
| CN111785329B (en) | Single-cell RNA sequencing clustering method based on countermeasure automatic encoder | |
| CN108171136A (en) | A kind of multitask bayonet vehicle is to scheme to search the system and method for figure | |
| CN111897733B (en) | Fuzzy test method and device based on minimum set coverage | |
| CN109977994B (en) | Representative image selection method based on multi-example active learning | |
| CN108805193B (en) | Electric power missing data filling method based on hybrid strategy | |
| CN108229588B (en) | Machine learning identification method based on deep learning | |
| CN109543727A (en) | A kind of semi-supervised method for detecting abnormality based on competition reconstruct study | |
| Zhong et al. | A comparative study of image classification algorithms for Foraminifera identification | |
| CN100416599C (en) | Not supervised classification process of artificial immunity in remote sensing images | |
| CN110751121A (en) | Unsupervised radar signal sorting method based on clustering and SOFM | |
| Tang et al. | A phylogenetic scan test on a Dirichlet-tree multinomial model for microbiome data | |
| CN1873660A (en) | Supervised classification process of artificial immunity in remote sensing images | |
| CN111429965A (en) | T cell receptor corresponding epitope prediction method based on multiconnector characteristics | |
| CN108229692B (en) | Machine learning identification method based on dual contrast learning | |
| CN104573004B (en) | A kind of double clustering methods of the gene expression data based on double rank genetic computations | |
| CN1873658A (en) | Method for selecting features of artificial immunity in remote sensing images | |
| CN1317677C (en) | Genetic algorithm based human face sample generating method | |
| CN108229693B (en) | Machine learning identification device and method based on comparison learning | |
| CN116152644A (en) | Long-tail object identification method based on artificial synthetic data and multi-source transfer learning | |
| Hu et al. | A classification surrogate model based evolutionary algorithm for neural network structure learning | |
| CN110427973A (en) | A kind of classification method towards ambiguity tagging sample |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080409 Termination date: 20110629 |